AI consultant vs AI agency vs productized vs fractional AI Operations.

By the time someone books a call with me, they’ve already looked at the four options. They have a tab open for the AI agency, another for the productized service that promised to set up their AI bookkeeping in two weeks, and the LinkedIn page of someone whose title says AI consultant. They tell me they don’t know which is right for the work they’re trying to do. The four options are four different shapes of relationship, not four different prices for the same thing.

The short version

Four categories of help, all called AI something. An AI consultant is one person you hire for diagnosis or advice. Hourly or per-project. They tell you what to do, sometimes they build it. An AI agency is a team that delivers an AI project the same way a marketing agency delivers marketing. Scoped and billed by project, handed off when the deliverable lands. A productized AI service is a fixed-price offer that does one specific thing for many businesses, like setting up AI bookkeeping or running AI outbound for a sales team. If your need matches the template, it’s the cheapest route. A fractional AI Operations engagement is a senior operator embedded part-time. They scope, build the systems, and stay on as your team learns to run them. The right one depends on whether your operations are stable and need AI on top, or whether the operations themselves need to be rebuilt around AI.

What fractional AI Operations actually looks like inside

1. One senior operator doing all of it. Agencies have hierarchies: a salesperson finds you, an account manager schedules, junior devs build. Productized services have a setup window followed by a help desk. Consultants come alone but often won’t build. A fractional engagement is one senior person from first call through the build, and on through the monthly partnership if you want that next. The senior person is the same person on every call.
2. Operations language, not engineering language. AI agencies pitch in terms of models, APIs, prompt engineering, latency. A fractional engagement talks in workflows, handoffs, decision points, who owns what after the build. The vocabulary matches the buyer, who is operational.
3. Custom per engagement, not templates. Productized services scale by selling the same template to many businesses. Agencies tend toward repeatable project shapes too: a marketing automation project looks roughly the same client to client. Consultants recycle the same diagnostic frameworks. A fractional engagement is the shape your operation actually needs. If three companies hire me, the three plans come out different because their operations are different.
4. Investment is published. Scoping, build, and monthly partnership ranges all live on the investment page. Agencies quote privately, productized services list the headline price but not always the lock-in, consultants vary by reputation. Asking what something costs should not be the third email.

What it isn't

• It isn’t an agency without the team. I get asked this one early. The shape can look the same: scope, build, handoff. The answer is no for a structural reason. One senior person doing all of it, custom per engagement, no hierarchy of sales-to-builder handoffs, investment published. The work fits inside a project shape but the relationship doesn’t read as agency from the inside.
• It isn’t a consultant who happens to know AI. A consultant diagnoses and recommends. A fractional AI Operations role builds the thing and stays to operate it.
• It isn’t a productized service with custom paint. Productized means the same template for many companies. Fractional means whatever shape your operation actually needs. If the productized service does exactly what you need, take that route and save the money.
• It isn’t a fractional CTO. A fractional CTO owns technology decisions across the whole company. A fractional AI Operations Architect owns the AI shape of the operations layer. Adjacent, but the work is different.

How to know which to hire

1. Hire an AI consultant when you need a diagnosis more than a deliverable. You want a 60-minute conversation that points you in the right direction, plus a writeup. You don’t need someone to do the building.
2. Hire an AI agency when your operations are stable and you want a specific AI project delivered by a team. Marketing automation, a custom dashboard, or an AI feature inside the product you’re building. You know what you want. You want it scoped, delivered, off your desk.
3. Buy a productized AI service when your need is exactly what they sell. AI bookkeeping setup is the canonical example, but there are productized offers now for AI outbound, AI customer service, lead scoring, even AI scheduling for service businesses. If the product fits, you save money and time. If you need a custom version of their thing, the savings disappear and you should have gone fractional from the start.
4. Hire a fractional AI Operations Architect when the operations themselves need to change shape. Your team is drowning in work AI could absorb, your workflows were built before AI was a thing you could buy, and the operation needs rebuilding, not a few new tools bolted onto the side. When someone tells me “we tried Notion AI and it didn’t change anything,” it’s usually this category. The tool worked. The operation around it didn’t.

Why this category is new

Three years ago, this role didn’t exist. AI wasn’t useful enough yet for anyone to hire someone to wire it into a business. The workflows small businesses ran on were stable.

AI became operationally useful in the last eighteen months, and the workflows that had been stable for a decade started having an AI version that could absorb half the manual work in them. Designing that shape is its own work, neither engineering nor consulting. The hardest part isn’t the model and it isn’t the advice. It can’t be productized either, because every operation has a different shape. What fits is one operator who can scope, build, and stay.

The role showed up because there was nothing else that worked. I ended up doing this because I’d already been running operations across different industries for a decade. The tools changed. The work didn’t.

Here's what I do

I run fractional AI Operations engagements out of Portland. Two weeks of scoping, then a build that runs three to six months, then a monthly partnership for clients who want one after. The work I take is the rebuild-the-operations kind, not the AI-bolted-onto-the-side kind.

If you’re not sure which of the four you need, that’s what the Scoping Engagement is for. Two weeks of looking, then a written plan with three things worth building, each with an investment range and a timeline. If you want the version of this question shaped for your operation, start a conversation.